知识卡片
训练目标低不等于现实错误低
训练样本只是现实分布的有限近似,最小化训练目标可能学习样本偏差或噪声。监控留出数据和不同群组、时间条件有助识别问题,但仍不能保证未来;扩大网络或使训练损失归零也不自动改善应用。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 神经网络基础
原书出处 · 1 处
- · 查看原书依据
Another training approach is batch learning. A batch is a subset of the training data—for example, 0.1%, 1%, 10%, or 20% of your dataset. Each batch creates a new error surface to experiment with as you ski around searching for the unknown “global” error surface minimum. Your training data is just a sample of the examples that will occur in the real world, so your
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第05章《神经网络基础》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-5.html)
- 结论依据:摘录中关于训练目标低不等于现实错误低的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Another training approach is batch learning. A batch is a subset of the training data—for example, 0.1%, 1%, 10%, or 20% of your dataset. Each batch creates a new error surface to experiment with as you ski around searching for the unknown “global” error surface minimum. Your training data is just a sample of the examples that will occur in the real world, so your